Speaker Localization in Reverberant Noisy Environment Using Principal Eigenvector and Classifier

Abstract

Sound source localization is essential in many microphone arrays application, ranging from speech enhancement to human-computer interface. The steered response power (SRP) using the phase transform (SRP-PHAT) method has been proved robust, but the algorithm may fail to locate the sound source in highly reverberant noisy environment. The Naive-Bayes localization algorithm based on classification of cross-correlation functions outperforms the SRP-PHAT in highly reverberant noisy environment. This paper proposes the improved Naive-Bayes localization algorithm using principal eigenvector. Simulation results have demonstrated that the proposed algorithm provides higher localization accuracy than the Naive-Bayes algorithm in reverberant noisy environment.